Setting my WiFi adhoc connection in power save mode

Setting my WiFi adhoc connection in power save mode

[Jorge Miguel]

Hello,

I am trying to set power save mode in my Intel 3945ABG wireless card
under Ubuntu 10.10 - Maverick Meerkat and iwl3945 driver.

I need to set power save mode. In order to do so I tried to change
period and timeout options of iwconfig command so that beacon period
and ATIM window are changed.

For instance
:~$ sudo iwconfig wlan0 power period 2
    Error for wireless request "Set Power Management" (8B2C):
    SET failed on device wlan0 ; Invalid argument.

As you can see I get an error. Is it because it is not supported in my
card? Is there any way to set beacon interval and ATIM window
parameters?

Many thanks in advance,
Jorge

Re: Setting my WiFi adhoc connection in power save mode

[Stanislaw Gruszka]

On Thu, May 12, 2011 at 07:02:23PM +0200, Jorge Miguel wrote:
> I am trying to set power save mode in my Intel 3945ABG wireless card
> under Ubuntu 10.10 - Maverick Meerkat and iwl3945 driver.
> 
> I need to set power save mode. In order to do so I tried to change
> period and timeout options of iwconfig command so that beacon period
> and ATIM window are changed.
> 
> For instance
> :~$ sudo iwconfig wlan0 power period 2
>     Error for wireless request "Set Power Management" (8B2C):
>     SET failed on device wlan0 ; Invalid argument.
> 
> As you can see I get an error. Is it because it is not supported in my
> card? Is there any way to set beacon interval and ATIM window
> parameters?

This should work until you do this in managed mode, power save is only a
infrastructure network feature. Particularly this will not work on IBSS
network  (ad-hoc mode).

Also in kernel 2.6.39, after driver split into iwlegacy, power save
is marked as not supported by driver. I wonder it was deliberate action
from Intel developers, or just IEEE80211_HW_SUPPORTS_PS is missed by
mistake. I'm going to look at this.

Stanislaw

Re: Setting my WiFi adhoc connection in power save mode

[Johannes Berg]

On Fri, 2011-05-13 at 13:50 +0200, Stanislaw Gruszka wrote:

> > I need to set power save mode. In order to do so I tried to change
> > period and timeout options of iwconfig command so that beacon period
> > and ATIM window are changed.
> > 
> > For instance
> > :~$ sudo iwconfig wlan0 power period 2
> >     Error for wireless request "Set Power Management" (8B2C):
> >     SET failed on device wlan0 ; Invalid argument.
> > 
> > As you can see I get an error. Is it because it is not supported in my
> > card? Is there any way to set beacon interval and ATIM window
> > parameters?
> 
> This should work until you do this in managed mode, power save is only a
> infrastructure network feature. Particularly this will not work on IBSS
> network  (ad-hoc mode).

Technically there is a PS mechanism for IBSS (ATIM and all that) but
it's not supported in the stack right now.

> Also in kernel 2.6.39, after driver split into iwlegacy, power save
> is marked as not supported by driver. I wonder it was deliberate action
> from Intel developers, or just IEEE80211_HW_SUPPORTS_PS is missed by
> mistake. I'm going to look at this.

Deliberate. On some platforms/systems, the way the device goes to sleep
caused DMA trouble or something like that, and the ucode would sysassert
when it tried to wake itself up. The problem was fairly rare, but of
course the device wakes up 10 times a second or so ...

johannes

Re: Setting my WiFi adhoc connection in power save mode

[Stanislaw Gruszka]

On Fri, May 13, 2011 at 01:55:03PM +0200, Johannes Berg wrote:
> > Also in kernel 2.6.39, after driver split into iwlegacy, power save
> > is marked as not supported by driver. I wonder it was deliberate action
> > from Intel developers, or just IEEE80211_HW_SUPPORTS_PS is missed by
> > mistake. I'm going to look at this.
> 
> Deliberate. On some platforms/systems, the way the device goes to sleep
> caused DMA trouble or something like that, and the ucode would sysassert
> when it tried to wake itself up. The problem was fairly rare, but of
> course the device wakes up 10 times a second or so ...

It's interesting if PS on 3945 ever works or was broken from very
beginning. I bet it was, on some old kernel, and this issue is similar
case like most of other bugs we have in iwlwifi - something that worked
well was broken by adding new chipsets/features support, but never fixed
again and just masked/disabled instead.
 
Stanislaw

Re: Setting my WiFi adhoc connection in power save mode

[Johannes Berg]

On Fri, 2011-05-13 at 14:22 +0200, Stanislaw Gruszka wrote:
> On Fri, May 13, 2011 at 01:55:03PM +0200, Johannes Berg wrote:
> > > Also in kernel 2.6.39, after driver split into iwlegacy, power save
> > > is marked as not supported by driver. I wonder it was deliberate action
> > > from Intel developers, or just IEEE80211_HW_SUPPORTS_PS is missed by
> > > mistake. I'm going to look at this.
> > 
> > Deliberate. On some platforms/systems, the way the device goes to sleep
> > caused DMA trouble or something like that, and the ucode would sysassert
> > when it tried to wake itself up. The problem was fairly rare, but of
> > course the device wakes up 10 times a second or so ...
> 
> It's interesting if PS on 3945 ever works or was broken from very
> beginning. I bet it was, on some old kernel, and this issue is similar
> case like most of other bugs we have in iwlwifi - something that worked
> well was broken by adding new chipsets/features support, but never fixed
> again and just masked/disabled instead.

I doubt it, iirc it was just some DMA issue that was discovered at some
point and nobody could really pinpoint it. Those old devices stored some
internal data into a host memory buffer, and reading them back sometimes
corrupted them.

johannes

Re: Setting my WiFi adhoc connection in power save mode

[Stanislaw Gruszka]

On Fri, May 13, 2011 at 02:27:22PM +0200, Johannes Berg wrote:
> > > Deliberate. On some platforms/systems, the way the device goes to sleep
> > > caused DMA trouble or something like that, and the ucode would sysassert
> > > when it tried to wake itself up. The problem was fairly rare, but of
> > > course the device wakes up 10 times a second or so ...
> > 
> > It's interesting if PS on 3945 ever works or was broken from very
> > beginning. I bet it was, on some old kernel, and this issue is similar
> > case like most of other bugs we have in iwlwifi - something that worked
> > well was broken by adding new chipsets/features support, but never fixed
> > again and just masked/disabled instead.
> 
> I doubt it, iirc it was just some DMA issue that was discovered at some
> point and nobody could really pinpoint it. Those old devices stored some
> internal data into a host memory buffer, and reading them back sometimes
> corrupted them.

Ok, so I will probably not try to reenable PS (as long someone clearly
state that it works in the past, and this is a regression).

Does this problem happen on both 4965 and 3945 ?

Stanislaw

Re: Setting my WiFi adhoc connection in power save mode

[Johannes Berg]

On Fri, 2011-05-13 at 14:39 +0200, Stanislaw Gruszka wrote:

> > I doubt it, iirc it was just some DMA issue that was discovered at some
> > point and nobody could really pinpoint it. Those old devices stored some
> > internal data into a host memory buffer, and reading them back sometimes
> > corrupted them.
> 
> Ok, so I will probably not try to reenable PS (as long someone clearly
> state that it works in the past, and this is a regression).

Well, problem is that it _did_ work for some people, since it seemed to
depend on the platform. I don't remember all the details though.

> Does this problem happen on both 4965 and 3945 ?

I'm not sure on 3945. ISTR there was an issue, but I don't know if it
was the same one.

johannes

Re: Setting my WiFi adhoc connection in power save mode

[Jorge Miguel]

Hi,

I am really glad I got all your answers. Thanks a lot. Now it is clear to me.
Jorge.



On 13 May 2011 14:45, Johannes Berg <johannes@sipsolutions.net> wrote:
> On Fri, 2011-05-13 at 14:39 +0200, Stanislaw Gruszka wrote:
>
>> > I doubt it, iirc it was just some DMA issue that was discovered at some
>> > point and nobody could really pinpoint it. Those old devices stored some
>> > internal data into a host memory buffer, and reading them back sometimes
>> > corrupted them.
>>
>> Ok, so I will probably not try to reenable PS (as long someone clearly
>> state that it works in the past, and this is a regression).
>
> Well, problem is that it _did_ work for some people, since it seemed to
> depend on the platform. I don't remember all the details though.
>
>> Does this problem happen on both 4965 and 3945 ?
>
> I'm not sure on 3945. ISTR there was an issue, but I don't know if it
> was the same one.
>
> johannes
>
>